JP2004206608A - Document retrieval method, its device, and its program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document retrieval method capable of improving retrieval precision for document retrieval from a retrieval keyword inputted by a user and improving convenience based on automatic recommendation of a next limitted retrieval candidate. <P>SOLUTION: In this document retrieval method using inclusion of all the keywords designated by a user as a retrieval condition, a document satisfying a condition requiring correspondence of both factors, which are character strings of the keywords designated by the user and appearance sequence of the keywords, is retrieved. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a document search method, apparatus, and program for searching a large number of documents for documents that match conditions specified by a user.
[0002]
[Prior art]
In the conventional keyword search, first, as a preprocessing of the search, all the character strings appearing in one document are extracted as a keyword for each document to be searched, and this keyword is used as a keyword for searching this document. It is stored as index data for each group. Then, the index data is compared with which document contains the keyword input by the user, and the corresponding document is displayed as a search result for the document group.
[0003]
However, there is no search method in which a search condition is a match between the order in which a user inputs a plurality of keywords and the order in which keywords appear in a document.
[0004]
Further, in a structure description language (eg, HTML), a document is divided into a plurality of blocks in a document, and the contents described for each block may be significantly different. There is no search method for specifying a condition that all keywords entered by the user are included.
[0005]
As a function of recommending the next search candidate to the user, a method is conventionally disclosed in Patent Document 1 titled “Next search candidate word presentation method and apparatus and recording medium recording next search candidate word presentation program”. There has been a method of recommending, as a search keyword candidate, a keyword that has been frequently input by a past user.
[0006]
In addition, using the method disclosed in “Study of Automatic Text Classification System for Questionnaire” in Non-Patent Document 1, a document group of search results is classified by a character string, and the classification result is shown to a user. There was also a method of recommending search candidates.
[0007]
However, there is a problem that accuracy is poor because none of the above reflects the document structure existing in the search target document.
[0008]
[Patent Document 1]
JP-A-2002-92032 [Non-Patent Document 1]
Masayuki Sugizaki, Masakatsu Okubo, Kazuo Tanaka, "Study of Automatic Text Classification System for Questionnaire", Information Processing Society of Japan, 59th (Late 1999) National Convention, 4N-3, Volume 2, pp. 377-378 (1999).
[0009]
[Problems to be solved by the invention]
The present invention has been made in view of the above circumstances, and provides a document search method capable of improving the search accuracy of a document search from a search keyword input by a user and improving the convenience by automatically recommending the next narrowed search candidate. It is intended to provide a device and a program.
[0010]
[Means for Solving the Problems]
In order to achieve the above object, the present invention provides a document search method using a search condition that includes all of a plurality of keywords specified by a user. A step of searching for a document that satisfies a condition that both elements match.
[0011]
In addition, the document search method of the present invention includes a step of searching for a document including a keyword specified by a user, and a step of searching for a keyword that narrows the search result by using a character string appearing after the keyword in each document of the search result. Automatically recommending to the user.
[0012]
The document search method according to the present invention includes a step of recognizing a plurality of document structure blocks in the search target document and a condition that the block includes all of a plurality of keyword character strings specified by a user. Searching for a document that satisfies the condition.
[0013]
Further, according to the present invention, in the document search method, a keyword that further narrows down a document search result by using a character string that appears in a block after a document structure block including all of a plurality of keyword character strings specified by a user is used. The method has a step of automatically recommending to a user.
[0014]
Further, the present invention is a document search apparatus using a search condition that includes all of a plurality of keywords specified by a user, wherein both elements of a character string of the keyword specified by the user and an appearance order thereof match. And a means for searching for a document that satisfies the condition.
[0015]
Further, the document search device of the present invention includes means for searching for a document including a keyword specified by a user, and a keyword for narrowing down the search result by using a character string appearing after the keyword in each document of the search result. Means for automatically recommending to a user.
[0016]
In addition, the document search device of the present invention includes means for recognizing a plurality of document structure blocks present in a search target document, and a condition that a plurality of keyword character strings specified by a user are all included in the block. Means for searching for a document that satisfies the condition.
[0017]
Further, according to the present invention, in the document search device, a keyword that further narrows down a document search result by using a character string appearing in a block after a document structure block including all of a plurality of keyword character strings specified by a user. A feature is provided that means for automatically recommending to a user is provided.
[0018]
The present invention is also a document search program that uses a search condition that includes all of a plurality of keywords specified by a user, wherein both elements of a character string of the keyword specified by the user and an appearance order thereof match. This is for causing a computer to execute a procedure for searching for a document satisfying the condition.
[0019]
Further, the document search program of the present invention automatically searches for a document that includes a keyword specified by a user, and automatically narrows down a search result by using a character string appearing after the keyword in each document of the search result. To make the computer execute a procedure recommended to the user.
[0020]
The document search program according to the present invention includes a procedure for recognizing a plurality of document structure blocks in a search target document, and a document satisfying a condition that all of a plurality of keyword character strings specified by a user are included in the block. This is for causing a computer to execute a procedure for searching for.
[0021]
Further, according to the present invention, in the document search program, a keyword that further narrows down a document search result by using a character string that appears in a block after a document structure block including all of a plurality of keyword character strings specified by a user is used. This is for making a computer automatically execute a procedure recommended to a user.
[0022]
That is, the order of appearance of the keywords and the document block as the document structure in the document to be searched are considered as the search condition together with the search condition from the user.
[0023]
Regarding the order of appearance of keywords, when extracting keywords that appear in a document for each document at the time of index data creation, information on the order of appearance of the keywords in the document is also associated with the keywords and indexed for each keyword. Keep in the data. Then, when the user searches for a document, the character strings and the order of the plurality of keywords input by the user are compared with the index data, and the keywords all match, and the condition that the magnitude relation of the order is correct matches. Let things be search results.
[0024]
Regarding document blocks, when extracting keywords that appear in the documents for each document at the time of index data creation, first analyze the structure of the documents and block them, and then attach an ID (identification) to each of them. The ID is associated with the keyword and is stored in the index data for each keyword. Then, when the user searches for a document, the character string of a plurality of keywords input by the user and the block ID are compared with the index data, and all keywords match and the condition that the block ID is the same is matched. Is the search result.
[0025]
Also, using a document block, a keyword character string included in a block ID located later in the document than the block ID to which the keyword belongs in the search result document is extracted from the index file. Recommended as a refined search keyword candidate.
[0026]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0027]
The document search device according to the embodiment of the present invention can be realized by, for example, the configuration of FIG. FIG. 1 shows two document search devices, “(1) document search device (index data generation)” and “(2) document search device (document search)”, which are implemented in the same device. It is also possible to do this. Each of the configurations in FIG. 1 will be described.
[0028]
(1) Document search device (index data generation)
A, a search target document group This is a database that stores a document group to be searched. It is not necessary for the database to have a proper search function. It is only necessary to store the search target documents in a configuration that can be retrieved separately for each document, so relatively simple storage such as a computer file system It can also be realized with functions.
[0029]
B: Access the search target document group database of the document selection A, retrieve the document data one by one from the database in order, and pass the document data to the processing to be performed thereafter.
[0030]
C. A process of extracting a plurality of search keywords from the document is performed on the document data extracted by the document selection of the keyword extraction B by an existing method such as natural language processing.
[0031]
D, Document structure extraction An operation of extracting the structure of the document is performed. The operation of extracting the structure of the document will be described later in detail.
[0032]
E. The keywords and the document structure extracted by the keyword extraction of the index data C and the document structure extraction of the index data D are stored in a database for each document in a format that can be easily searched at the time of search.
[0033]
(2) Document search device (document search)
a, keyword extraction A keyword is extracted from the character string input by the user with the input device of g. In keyword extraction, keywords are ordered according to the order in which the user inputs character strings.
[0034]
As a user interface provided to the user, for example, a screen for the user to input a keyword as shown in FIG. 2 is used. The user inputs a character string directly from the input device of g into the character string input space. If the user requests to enter a plurality of keywords, the character strings are entered by separating them with spaces.
[0035]
Natural language processing is applied to each character string in the same manner as in the above-described keyword extraction technique C to extract a plurality of keywords from the character string. As shown in FIG. 2, in “Today's weather”, two keywords “today” and “weather” can be extracted by general natural language processing.
[0036]
When extracting keywords, the extraction order, that is, the order from the beginning of each keyword is also extracted, and the order is associated with each keyword.
[0037]
b, Search condition generation A search condition for searching the index data of e is generated based on the plurality of keywords obtained by the keyword extraction of a and the order of each keyword. The search conditions will be described later in detail.
[0038]
c, access the index data of the search execution e, and search by the search condition generated by the search condition generation of b.
[0039]
d. The search result obtained by executing the search of the search result display c is presented to the user. In addition, a function of automatically presenting a keyword necessary for performing a further refined search on the obtained search result is also realized here. The recommendation of the refined search keyword will be described later in detail.
[0040]
e, a database in which a keyword and a document structure for each document obtained from the index data of the index data E are stored in a format that is easy to search at the time of search.
[0041]
f, a display device that presents information obtained by the search result display of the display device d to the user, and can be realized by a display device of a computer.
[0042]
g, input device This is an input device for the user to input search conditions as a character string. This can be realized by a computer keyboard device.
[0043]
[Extraction of document structure]
The document structure extraction of D will be described in detail.
[0044]
There are the following two document structures.
[0045]
D-1) Order of appearance of keywords in document D-2) Document structure block D-1 in the document uses the order of appearance of keywords in the document. When keywords are extracted from a document by natural language processing, the order in which the keywords are extracted, that is, the order of the extracted keywords from the head of the document is also calculated, and the extracted keywords are associated with the extracted keywords and stored in the index data of E.
[0046]
D-2 first recognizes a document structure block from the document. FIG. 3 is used to explain the document structure block. FIG. 3 shows HTML (hypertext markup language) used in a general web page.
[0047]
In this HTML, one piece of information is included between <td> and </ td>, and when it is displayed by an HTML browser, it is displayed in a table format as shown in FIG. FIG. 4 shows an example in which the web page of FIG. 3 is displayed by a web browser. In the present embodiment, such a range from <td> to </ td> is used for a search as a document structure block. That is, when <td> to </ td> are recognized as one block, IDs are assigned to them, and a keyword is extracted from the document. When extracting the keyword from the document, the index data of E is associated with the extracted keyword of the document structure block ID. To hold.
[0048]
HTML is a structure description language, and various HTML tags exist in a document. Therefore, it is possible to recognize the document structure block not only from <td> to </ td> but also in other HTML tags. In addition to HTML, in a general Japanese document, it is common to add "." To the end of a sentence, but by using this, from immediately after "." To the next "." Can be recognized as a document structure block.
[0049]
[Generate search condition]
The generation of the search condition b will be described in detail. There are the following two methods for generating search conditions depending on the extraction of the document structure of D.
[0050]
b-1) Match the appearance order of keywords. When the document structure extraction is D-1 b-2) Matching of document structure blocks. When the document structure extraction is D-2, regarding the matching of the appearance order of the keyword of b-1, the appearance order of the keyword extracted from the character string input at the time of the search by the user and the appearance order of each document extracted from the search target document Matching of the appearance order of keywords is set as a search condition. For example, when the user inputs the keywords “today” and “weather”, only documents in which the keywords appear in the order of “today” and “weather” are searched. Documents in which keywords appear in the order of “weather” and “today” are not searched.
[0051]
Regarding the match of the document structure block of b-2, the search condition is that the keyword extracted from the character string input at the time of the search belongs to the same document structure block in the search target document. I do. For example, when the user inputs the keywords “today” and “weather”, only documents having the same document structure block ID as “today” and “weather” are searched.
[0052]
[Recommended search keywords]
The search result display of d will be described in detail. When displaying search results, there is a user's need to further narrow down the search results and to narrow down and display search result candidates. Therefore, a function of automatically presenting narrowed-down keyword candidates is useful. By utilizing the document structure according to the embodiment, an appropriate narrowing keyword can be presented.
[0053]
There are the following two methods of presenting the refined search keyword depending on the extraction of the document structure of D.
[0054]
d-1) Use of the appearance order of keywords. When document structure extraction is D-1 d-2) Use of document structure block. In the case where the document structure extraction is D-2, the use of the keyword appearance order of d-1 can be realized when the prior document structure extraction utilizes the order of appearance of the keywords in the document.
[0055]
At the time of search, first, the character string that hits the keyword used for the search is specified in which position in each document to be searched. Then, narrow-down keyword candidates are extracted from character strings that appear later in the document than that position.
[0056]
In the search target document, the user can arbitrarily select whether to extract a narrowing keyword from the second half of the character string that hits the search or extract it from the first half. A configuration is also possible.
[0057]
The use of the document structure block d-2 can be realized when the document structure extraction in advance uses the document structure block in the document.
[0058]
At the time of retrieval, a document in which a plurality of keywords entered by the user as in b-2 above belongs to the same block in the document is presented to the user. Then, recommendation of the narrowed-down keyword candidates to the user can be realized by further extracting from the same block as the keyword inputted by the user.
[0059]
The above-described embodiment relates to a method of searching a document group stored on the document search apparatus by utilizing the document structure extraction method of the embodiment.
[0060]
However, the present embodiment is effective not only for a document group stored on a device, but also for a search for a document group distributed and stored in various devices. This embodiment will be described below.
[0061]
The (1) index device shown in FIG. 5 has a function of extracting a document structure and a keyword from a group of documents on the Internet, storing them in an index file, and preparing for a search request from the document search device. (2) The document search device has a function of making a search request to the index device based on a keyword input by the user from the input device, and presenting the result to the user.
[0062]
(1) Indexing device The retrieval target document (A) is continuously read from a plurality of servers on the Internet. Documents to be searched are listed by the document selection (B) function, and the documents are continuously read out based on the list. When the initial document is set, the list is used to know the existence of other documents by using the relationship between the documents in the initial document, add the document to the list, and further include it in those documents There is a recursive method of adding another document to a list based on the relation to another document. In some cases, all lists are created manually, and only fixed documents are searched. There is also a method of recursively creating a list for a certain range of documents by mixing these two methods.
[0063]
The search target document read by the document selection (B) function is subjected to a keyword extraction (C) function to extract a plurality of search keywords from the document by an existing method such as natural language processing.
[0064]
Further, an operation of extracting a document structure is performed by a document structure extraction (D) function. There are two types of document structure extraction methods, one based on the order of appearance of keywords in a document (method 1) and the other based on document structure blocks in a document (method 2).
[0065]
The former uses the order of appearance of keywords in a document. When keywords are extracted from a document by natural language processing, the order in which the keywords are extracted, that is, the order of the extracted keywords from the head of the document is also calculated, and the extracted keywords are associated with the extracted keywords and stored in the index data (E).
[0066]
The latter first recognizes a document structure block from the document. Most of the documents existing on the Internet are structured description languages such as html and xml, and the structured description language is structured by tags. This structure is recognized as one block, and IDs are assigned to the blocks to distinguish them. When extracting a keyword from a document, the document structure block ID is held in the index data (E) in association with the extracted keyword.
[0067]
In addition to html and xml, in general Japanese documents, it is common to add "." To the end of a sentence, and by using this, the next "." Immediately after "." Can be recognized as a document structure block. In addition, it is also possible to divide the meaning before and after using a connecting word used when the meaning of a document changes in Japanese, such as "for example" or "but", to form a document structure.
[0068]
The keywords and the document structure extracted by the keyword extraction (C) function and the document structure extraction (D) function are stored in a database for each document in a format that is easy to read at the time of retrieval.
[0069]
(2) Document Retrieval Device First, a keyword is extracted from the keyword input by the user from the input device by the keyword extraction (a) function. A keyword is extracted for each character string input by the user. When the user inputs “Today's weather”, for example, two keywords “today” and “weather” are extracted. Not only keywords but also the order of keywords is kept.
[0070]
Next, based on the plurality of keywords obtained by the keyword extraction (a) function and the order of each keyword, a search condition for searching index data is generated by the search execution (c) function. There are the following two methods depending on the method of generating index data in (1).
[0071]
Match the order in which keywords appear. When the document structure extraction is the method 1, the document structure blocks are matched. When the document structure extraction is the second method.
[0072]
Regarding the matching of the order of appearance of keywords, it is determined that the order of appearance of keywords extracted from the character string input by the user at the time of search matches the order of appearance of keywords for each document extracted from the search target document. Search condition. For example, when the user inputs the keywords “today” and “weather”, only documents in which the keywords appear in the order of “today” and “weather” are searched. Documents in which keywords appear in the order of “weather” and “today” are not searched.
[0073]
As for the matching of the document structure blocks, a search is performed using a search condition that a keyword extracted from a character string input by the user at the time of search belongs to the same document structure block in the search target document. For example, when the user inputs the keywords “today” and “weather”, only documents having the same document structure block ID as “today” and “weather” are searched.
[0074]
According to the created search condition, a search request is issued for the index data (E) of the index device of (1). The index device of (1) returns a search result corresponding to the search request to the document search device of (2).
[0075]
This search result is presented to the user by a search result display (d) function.
[0076]
At this time, a function of automatically presenting a keyword necessary for performing a further refined search on the obtained search result is also realized here.
[0077]
There are two methods for automatically presenting keywords for a narrow search, depending on the document structure extraction method (1) using the index device. Use of the order of appearance of keywords and use of document structure blocks.
[0078]
The use of the order of appearance of the keywords is performed when the document structure extraction method in (1) is one. This can be realized when the prior document structure extraction uses the order of appearance of keywords in the document.
[0079]
At the time of search, first, the character string that hits the keyword used for the search is specified in which position in each document to be searched. Then, narrow-down keyword candidates are extracted from character strings that appear later in the document than that position.
[0080]
In the search target document, the user can arbitrarily select whether to extract a narrowing keyword from the second half of the character string that hits the search or extract it from the first half. A configuration is also possible.
[0081]
The document structure block is used when the document structure extraction method in (1) is 2. This can be realized when the document structure extraction in advance uses a document structure block in the document.
[0082]
At the time of retrieval, a document in which a plurality of keywords entered by the user as in b-2 above belongs to the same block in the document is presented to the user. Then, recommendation of the narrowed-down keyword candidates to the user can be realized by further extracting from the same block as the keyword inputted by the user.
[0083]
It should be noted that the document search method in the embodiment is specifically executed by a computer such as a personal computer based on a predetermined document search program in advance. The document search program can be recorded on a predetermined computer-readable recording medium such as a CD.
[0084]
【The invention's effect】
As described above, according to the present invention, when narrowing down a document by a keyword search, it is possible to use information of meaning linkage in the document, so that the event of (2) occurs after the event of (1). Since the search can be distinguished from the fact that the event (1) has occurred after the event (2), the search result desired by the user can be easily obtained.
[0085]
Conventionally, when recommending the next narrowing candidate to the user, a character string that can be narrowed down is extracted from the entire document, so that the candidate becomes ambiguous and the number of candidates is enormous, making it difficult to use. Thus, since the candidates are limited and the amount is also limited, a highly convenient service is possible.
[Brief description of the drawings]
FIG. 1 is a configuration explanatory diagram showing a document search device according to an embodiment of the present invention.
FIG. 2 is an explanatory diagram showing an example of a screen on which a user inputs a keyword according to the embodiment of the present invention.
FIG. 3 is an explanatory diagram showing an example of HTML used in a general web page.
FIG. 4 is an explanatory diagram showing an example in which the web page of FIG. 3 is displayed on a web browser.
FIG. 5 is an explanatory diagram showing an example of a configuration of a document search on the Internet according to the embodiment of the present invention.
[Explanation of symbols]
A: document group to be searched, B: document selection, C: keyword extraction, D: document structure extraction, E: index data, a: keyword extraction, b: search condition generation, c: search execution, d: search result display, e: index data, f: display device, g: input device.

Claims (12)

A document search method using a search condition that includes all of a plurality of keywords specified by a user,
A document search method comprising a step of searching for a document that satisfies a condition that both a character string of a keyword specified by a user and an element in the order of appearance match each other. Retrieving documents containing the keyword specified by the user;
Automatically recommending to a user a keyword for narrowing down the search result using a character string appearing after the keyword in each document of the search result. Recognizing a plurality of document structure blocks in the search target document;
Searching for a document that satisfies the condition that all of the plurality of keyword character strings specified by the user are included in the block. The document search method according to claim 3,
A step of automatically recommending to the user a keyword for further narrowing down the document search result by using a character string appearing in a block after the document structure block including all of the plurality of keyword character strings specified by the user A document search method characterized in that: A document search device that includes a search condition that includes all of a plurality of keywords specified by a user,
A document search apparatus comprising: means for searching for a document that satisfies a condition that both a character string of a keyword specified by a user and an element in the order of appearance match. Means for searching for a document containing the keyword specified by the user; means for automatically recommending to the user a keyword for narrowing down the search result by using a character string appearing after the keyword in each document of the search result. A document search device comprising: Means for recognizing a plurality of document structure blocks in the search target document;
Means for searching for a document that satisfies a condition that all of a plurality of keyword character strings specified by a user are included in the block. The document search device according to claim 7,
A means is provided for automatically recommending a keyword to further narrow down a document search result to a user by using a character string appearing in a block after a document structure block including all of a plurality of keyword character strings specified by a user. A document search device characterized by the following. A document search program using a search condition that includes all of a plurality of keywords specified by a user,
A document search program for causing a computer to execute a procedure for searching for a document that satisfies a condition that both a character string of a keyword specified by a user and an element in the order of appearance match. A computer includes a procedure for searching for a document including a keyword specified by a user, and a procedure for automatically recommending to a user a keyword for narrowing down a search result using a character string appearing after the keyword in each document of the search result. A document search program to be executed. A procedure for recognizing a plurality of document structure blocks in the search target document,
A document search program for causing a computer to execute a procedure for searching for a document that satisfies a condition that a plurality of keyword character strings specified by a user are all included in the block. The document search program according to claim 11,
Using a character string that appears in a block after a document structure block that includes all of a plurality of keyword character strings specified by a user, a computer automatically recommends a keyword for further narrowing down a document search result to a user. A document search program to be executed.

Images